Noor Azliza, Sabri (2004) Comparative Study Between Neural Network And Statistic In Handwritten Digit Recognition. Masters thesis, Universiti Utara Malaysia.
NOOR_AZLIZA_BT._SABRI.pdf
Restricted to Registered users only
Download (4MB) | Request a copy
1.NOOR_AZLIZA_BT._SABRI.pdf
Download (866kB) | Preview
Abstract
Neural network is one of the most Artificial Intelligent techniques. It has been implemented in various applications ranging from non technical applications to highly technical applications. The ability of neural network was originally inherited from statistical models such as regression. Handwritten recognition is one of the promising domains for neural network. Many studies have shown the success and efficacy of neural network in handwritten recognition. Yet, less study compares the performance of neural network and statistical method. Hence, this study aims to compare the generalization performance of neural network and statistical model in handwriting recognition domain. The results obtained are compared and presented in this paper. Multilayer Perceptron is chose as neural network model and Multiple Nonlinear Regression as statistic model. The result (percentage of correctness) indicated that neural network model is better in generalization than the statistic model. A total of 768 datasets was used for training. Neural network has produced a higher generalization value if compared to statistic which is 94.98% and 78.7% respectively.
Item Type: | Thesis (Masters) |
---|---|
Supervisor : | UNSPECIFIED |
Item ID: | 1382 |
Uncontrolled Keywords: | Nueral network, Artificial Intelligent, Statistic, Handwritten Digit Recognition |
Subjects: | T Technology > T Technology (General) |
Divisions: | Faculty and School System > Sekolah Siswazah |
Date Deposited: | 12 Jan 2010 03:10 |
Last Modified: | 24 Jul 2013 12:11 |
Department: | Sekolah Siswazah |
URI: | https://etd.uum.edu.my/id/eprint/1382 |